Identification of Accelerometer Nonwear Time and Sedentary Behavior

The use of objective monitoring, such as accelerometers, for quantifying physical activity (PA) and sedentary behaviors in public health research is now commonplace. Accelerometry is an objective, detailed, and sensitive measure of PA and is especially useful in quantifying time spent in various activity intensities (i.e., sedentary, light, moderate, vigorous). Yet, reducing accelerometer data to reliable, valid, meaningful, and transferable units is complicated and requires the researcher to make numerous decisions, for which there are different approaches and recommendations (e.g., see Esliger, Copeland, Barnes, & Tremblay, 2005; Ward, Evenson, Vaughn, Rodgers, & Troiano, 2005). Currently, there is no best practice for identifying periods when the participant does not wear the accelerometer (i.e., nonwear time) versus the time the participant engages in sedentary activities while wearing the unit. When there are extended consecutive epochs with zero counts, it is not possible to discern whether they result from sedentary behavior during accelerometer wear or if the unit was removed for a period of time (reasons may include forgetfulness, participation in contact or water sports, bathing, etc.). A limited number of studies have considered this issue, with inconsistent findings (Choi, Liu, Matthews, & Buchowski, 2010; Evenson & Terry, 2009; Song et al., 2010; Winkler, Gardiner, Healy, Clark, & Sugiyama, 2009). Combined, the results of these earlier studies suggest the most appropriate way to assess nonwear time may be to use a threshold of approximately 60–90 min of consecutive zero counts and allow up to 2 min of nonzero counts on either side. However, a critical gap in the existing research remains, which is the lack of evidence for using of this criterion. This conundrum has led researchers to apply a wide range of criteria to detect accelerometer nonwear time, generally comprising an arbitrary number of bouts of consecutive zero counts. These values have ranged from 10 (Eiberg et al., 2005) to 180 (van Coevering et al., 2005) consecutive minutes of zero counts. Applying such rules, however, may result in overestimating PA accumulation and underestimating sedentary time (or vice versa) and may provide an inaccurate assessment of participant compliance. Although logs and diaries may be used to record periods of nonwear and PA, reliance on self-report methodologies is not ideal due to issues with compliance, recall, and participant burden (Tudor-Locke & Myers, 2001). Thus, the primary aim of the current study was to investigate the accuracy of various automated rules for determining accelerometer nonwear time in a sample of predominantly desk-based office workers (using their self-reported nonwear times as a criterion). Second, we examined the effect of applying these rules to accelerometer data retention and classification of sedentary behaviors. Identification of Accelerometer Nonwear Time and Sedentary Behavior